Combat Mechanics? Attack Speed, Recovery

[Mysh]

@ peddroelm

 

I am now travelling overseas, I won't have access to PoE for 2 weeks, but have MATLAB and my scripts with me. If you can send me data, I can do the scripting.

 

@ iszathi

 

Please private message me your email and I can send you the scripts. Thanks!

 

Will be a slow horse for a couple of days.

[VaultDuke]

oh, didn't know the average defenses in PoE were that low, I can see now, why acc and DR aren't as important. sadly, this makes modals such as savage attack really no-brainers. there is no reason not to use it all the time. it thought it would at least be somewhat situational.

[Maverick87]

how much does a 2handed, heavy armor fighter benefit from lets say 10 to 18 dex?

[Matt516]

Things are not simple, for sure. I continually get different peaks for different conditions. I am currently confident though that dexterity builds get the edge on late game, while might builds yield higher DPS early. Let's keep some doubts though, till we verify the formulas.

 

Personally, I've got no doubts - damage modifiers can only continue to go up as the series progresses, which only helps Dex over Might. It's a certainty to me that at some point - whether it be in the PoE 1 expansion(s) or in PoE 2 - Might will be completely and totally overshadowed by Dex for all auto-attacker builds. Casters obviously still benefit from Might to get more out of each spell.

 

Looks like you've done some good work though - though I haven't checked your formulas over in detail. Looking forward to seeing what you can come up with. If you can produce some good graphs for communicating how the current system makes Might less and less relevant as the levels get higher, maybe we can convince Obsidian to change it - hopefully by just making Might multiplicative (the most obvious solution) or making weapon enchants actually change the base damage (the least intrusive and most intuitive solution).

[VaultDuke]

or they could just add high level talents like 3% extra damage per point of might...

[Matt516]

or they could just add high level talents like 3% extra damage per point of might...

 

Which is also an option, though a more difficult one to balance and apply than just making Might and weapon quality improvements work as players would (and should IMO) intuitively expect them to work. (i.e. both multiplicative with other damage modifiers)

 

Again, there are lots of possible solutions, and I'm sure Obsidian will implement some of them. But I'm gonna advocate for the ones I think make the most sense and are easiest to implement, all the same.

[illathid]

For what it's worth I think having weapon qualities change the base damage makes the most sense to me. But I'm also very cautious about about stacking multiplicative damage modifiers after the BB.

 

Another thought would be to change might to an integer. Not sure how the math would work out for that, but it's an idea.

[Matt516]

Another thought would be to change might to an integer. Not sure how the math would work out for that, but it's an idea.

 

That would favor DW fast weapons above all else.

 

Re: the danger of multiplicative modifiers - absolutely. That's why I'd only want to see Quality affect Base Damage and Might be multiplicative. In that order. Weapon quality is most important as it allows Might to scale into lategame in some way, isn't ludicrously impactful, and is just in general the most intuitive way to handle things. Might being multiplicative I'd like to see, but I can understand why it may not be a great idea.

 

I almost wonder if the quality not affecting base damage for calculating modifier damage thing is a bug. The base damage on the weapon does change in the info screen, after all. Bug or no bug, the current implementation is confusing as hell.

[famousringo]

 

Things are not simple, for sure. I continually get different peaks for different conditions. I am currently confident though that dexterity builds get the edge on late game, while might builds yield higher DPS early. Let's keep some doubts though, till we verify the formulas.

 

Personally, I've got no doubts - damage modifiers can only continue to go up as the series progresses, which only helps Dex over Might. It's a certainty to me that at some point - whether it be in the PoE 1 expansion(s) or in PoE 2 - Might will be completely and totally overshadowed by Dex for all auto-attacker builds. Casters obviously still benefit from Might to get more out of each spell.

 

Looks like you've done some good work though - though I haven't checked your formulas over in detail. Looking forward to seeing what you can come up with. If you can produce some good graphs for communicating how the current system makes Might less and less relevant as the levels get higher, maybe we can convince Obsidian to change it - hopefully by just making Might multiplicative (the most obvious solution) or making weapon enchants actually change the base damage (the least intrusive and most intuitive solution).

 

Even for casters, Might over Dex is not so clear cut. Getting more damage out of each spell is less important the more per encounter spells you have. Also, Might doesn't improve stacking of buffs or debuffs, but Dexterity lets you pile on those buffs faster or do damage faster, and get that spell off before a hit interrupts you, and react with that clutch heal or CC sooner.

 

Yep, kinda starting to wish I'd built my druid for dex even though it wasn't "recommended".

 

And yes, thank you to everybody digging deep on the math in PoE's incredibly opaque combat system. You're saving my sanity at the expense of your own.

[Kaylon]

Personally I find the warhammer to be the best one handed weapon. There's no monster in the game which is resistant to both crush and pierce.

[KKDragonLord]

 

okay, this time with locked framerate

 

Base

AT 52

REC 74

REL  98

 

Only Chant (Rec x1.2, REL x2)

AT 52

REC 60 -> 74*(1-0.2)=59.2

REL 51 = 98*(1-0.5)=49

 

Only Armor (Rec -50%)

AT 52  

Rec 112 -> 74*(1+0.5)=111

Rel 98

 

Only Dex (AT, REC, REL 24%)

AT 44  -> 52*(1-0.24)=39.52  <- wrong

REC 60 <- 74*(1-0.24)=56.24 <- wrong

REL 81 <- 98*(1-0.24)=74.48 <- wrong

 

Only Pen Shot (Rec -20%)

AT 52

REC 89 -> 74*(1+0.2)=88.8

REL 98

 

ALL

AT 44  -> 52*(1-0.24)=39.52  <- wrong

REC 90  -> 74*(1-0.24+0.2+0.5-0.2)=93.24  <- wrong

REL 43 -> 98*(1-0.5-0.24)=25.48  <- wrong, maybe multiplicative stacking?

 

Any suggestions?

 

The only way this can make any sense is if Dex isn't truly giving a 3% bonus to attack speed, but only a 2% bonus instead. This would mean  a 16% instead of a 24%. This can be happening because originally, Dex was meant to increase IAS by only 2%.

 

EDIT: Ooops, i was following the thread until a few days ago and didn't realize it had progressed far beyond this point being raised.

Edited April 15, 2015 by KKDragonLord

[voad1]

So does this mean duel wielding sabres or stilletos would be the way to go? Especially for a barbarian taking one stands alone? Then would vulnerable attack be something you want to stay away from? How improtant is accuracy? Is a weapon focus worth taking or should everyone just go for damage increasing talents and I guess the attack speed increase for dual wielding?

 

Also, should one dump might for dex? Say leave might 10 on a character, max dex and invest the rest elsewhere? Or should might still be maxed? And what armor is ideal? Heavier armor has more of an effect on dual wielding dps?

Edited April 15, 2015 by voad1

[Eyonik]

If you have a weapon with a bonus like crit modifier, does this bonus only count towards a hit with this specific weapon or to all hits?

I know you can wield 2 Hatches and get the Deflection bonus twice, but if you wield 2 Battle Axes (each has +0.5 crit bonus), do you get +0.5 crit bonus on each hit or +1 crit bonus?

 

Sorry if this question was already answered, but I couldn't find anything about it.

Edited April 16, 2015 by Eyonik

[voad1]

I think you could maybe find out by checking your character sheet and see if it gets surpressed. Unless that sheet is possibly inaccurate. I think it applies to just that weapon though. Just like how when dual wielding you have 2 seperate accuracy.

[Matt516]

So does this mean duel wielding sabres or stilletos would be the way to go? Especially for a barbarian taking one stands alone?

 

Hopefully OSA will be fixed soon, so Barbarian players won't be forced into dual wielding as a way to be optimal. :/

[b0rsuk]

 

 

What's the rate of fire for scepters ? The word "scepter" doesn't appear anywhere in this thread (until this point)! This is very important to me, I want to build a ranged monk. Also, scepter has better damage types than a rod. All tables and comparisons I can find (reddit, this forum) are without implements.

Edited April 16, 2015 by b0rsuk

[Kaylon]

Has someone checked using fraps if the modifiers stack additively for the reloading part of ranged weapons? I'm asking because when you look into Cheatengine they are multiplied (1.2x1.5x1.2 for a ranger with gunner, swift aim and chanter buff). 

OK, I tested myself using fraps/virtualdub how the reload time was calculated and everything is multiplicative like I suspected.

Here are the numbers for the arquebus (used the reload time without the delay you have between actions - which is about 5frames):

- 10dex, no other buff - 188frames (base)

- 10dex, swift aim (1.5 x reload time) - 126frames (188/1.5=125.3)

- 10dex, swift aim (1.5 x reload time), gunner (1.2 x reload time) - 105frames (188/1.5/1.2=104.4)

- 20dex, swift aim (1.5 x reload time), gunner (1.2 x reload time) - 81frames (188/1.5/1.2/1.3=80.3)

- 20dex, swift aim (1.5 x reload time), gunner (1.2 x reload time), chant (1.2 x reload time) - 66frames (188/1.5/1.2/1.3/1.2=66.95)

The results are pretty spot on, with the differences within 1frame.

[voad1]

 

 

So does this mean duel wielding sabres or stilletos would be the way to go? Especially for a barbarian taking one stands alone?

Hopefully OSA will be fixed soon, so Barbarian players won't be forced into dual wielding as a way to be optimal. :/

Lol, yes I suppose it is ridiculous but I am just trying to min max.

 

Maybe it was answered earlier here but is accuracy important enough to invest in a weapon focus For that +6 accuracy? Are dex and int noq the two key attributes for a barb and might can be safely ignored?

 

At -5 accuracy is savage attack worth taking foe the added 15% damage? At -20% speed is vulnerable attack worth taking for 5 dr bypass?

 

Does the recovery penalty of heavy armor greatly affect dps? Too many questions? Lol

[Matt516]

 

stuff

 

I don't have incredibly detailed math to back these answers up, mostly because "solid" (as in completely definitive) math on this kind of thing is dubious at best since there are tons of things that can vary (namely enemy Deflection and DR) that will change things. That said, I am familiar enough with the game mechanics and maths to give you some decent (I think) advice anyhow, with some ballpark calculations that I'm pretty comfortable with.

 

Re: weapon focus - Accuracy isn't that important, but if you're minmaxing for DPS you definitely want a weapon focus. You'll have room for it eventually.

 

Re: Might - You want to max Might (or close) - 20 Might is the same damage bonus you get from an Exceptional weapon. Not colossally important (especially as the series goes on and the weapon quality gets higher while Might stays the same, hopefully they fix that) but nothing to ignore either.

 

Re: Savage Attack - Absolutely yes. The most Accuracy can ever get you is ~1.5% average damage per point when it's converting misses to crits, though it usually gives you ~1% average damage per point when it's converting grazes to crits or misses to hits. That's ~7.5% max increased average damage from that 5 Accuracy. Whereas Savage Attack is 15%. So yeah - Savage Attack is absolutely worth it pretty much always.

 

Re: Vulnerable attack - My hunch (and it's a strong hunch) is yes. That will generally translate to a flat +5 damage against everything on each hit (few enemies have <5 DR), which is anywhere between ~45% of your weapon damage (fast weapons have mean damage of 11) to ~30% of your weapon damage (2H weapons have mean damage of 17). Or, in other words - 5 DR pen is about as good a bonus as having an Exceptional or Superb weapon in terms of damage. See why the Estoc is hideously OP now?

 

It's a little more complicated than the Accuracy<->Damage tradeoff due to an attack speed loss actually being a multiplicative loss (i.e. all your damage is multiplied by 0.8 or so) instead of an additive loss like Accuracy is (since graze/crit modifiers are additive). But my hunch is still that it's a good decision. Only one I'd be iffy about using it on is 2H weapons.

 

Re: Heavy armor - don't use heavy armor on a dps. Use the lightest armor you can put them in while still having them survive. That last part is important. Lots of people put glass cannon Rogues in clothes and then get all confused when they die when someone sneezes on them. Armor up as needed. My Barbarian wears Hide.

Edited April 17, 2015 by Matt516

[voad1]

Thanks a ton Matt, looks like I will be re rolling my barb again, lol. Went 10 10 18 10 18 12 or something like that last time around. Think I will use perception as a dump stat and pump up might.

 

Was looking at bloody slaughter but I think that has fallen lower on my list of talents to get. I was looking at taking accurate carnage and maybe barbaric blow. Might drop barbaric blow now, not certain. That leaves me 4 to 5 talents to take. I definitely want envenomed strike for it's raw dot. I think that mixed with carnage would be great. Also using a blunderbus as a secondary to open on boss fights would work well I think.

 

Sabres or Stilletos duel wielded. Take duel wield attack speed talent. An vulnerable attack. Negating the speed penalty and adding 5 dr bypass to each weapon. Savage attack and ruffian focus. Plus 6 accuracy to all weapons I plan on using stiletto, sabre, blunderbus. And +15 damage... kind of negating the penalty there as well. I may drop barbaric blow for the weapon focus or vice versa. Not sure.

 

Thought occured to me that with bloody savagery and barbaric blow you could make a barb that focus on weapons with crit procs such as knocks enemies prone crit proc. Use barbaric blow or when they get low health you'd hopefully be knocking mobs off their feet with carnage. Could be fun. But I am taking this off topic I think, sorry.

[Matt516]

Thanks a ton Matt, looks like I will be re rolling my barb again, lol. Went 10 10 18 10 18 12 or something like that last time around. Think I will use perception as a dump stat and pump up might.

 

Was looking at bloody slaughter but I think that has fallen lower on my list of talents to get. I was looking at taking accurate carnage and maybe barbaric blow. Might drop barbaric blow now, not certain. That leaves me 4 to 5 talents to take. I definitely want envenomed strike for it's raw dot. I think that mixed with carnage would be great. Also using a blunderbus as a secondary to open on boss fights would work well I think.

 

Sabres or Stilletos duel wielded. Take duel wield attack speed talent. An vulnerable attack. Negating the speed penalty and adding 5 dr bypass to each weapon. Savage attack and ruffian focus. Plus 6 accuracy to all weapons I plan on using stiletto, sabre, blunderbus. And +15 damage... kind of negating the penalty there as well. I may drop barbaric blow for the weapon focus or vice versa. Not sure.

 

Thought occured to me that with bloody savagery and barbaric blow you could make a barb that focus on weapons with crit procs such as knocks enemies prone crit proc. Use barbaric blow or when they get low health you'd hopefully be knocking mobs off their feet with carnage. Could be fun. But I am taking this off topic I think, sorry.

 

My Barb is 16/10/10/8/16/18 - not optimal for DPS of course, but I like the Resolve dialogue options and can do some decent and high AoE damage - plus I almost never get interrupted. And I use a Greatsword. It's a weirdish build but I'm enjoying it - and it's certainly viable (at least on Hard).

 

If you're looking for more abilities and talents to burn, there's always the "Brute Force" (I think that's what its called) and "Threatening Presence" combo. Attack Fortitude if it's the lower defense, and automatically AoE debuff enemies' Fortitude at the same time. It's fun. Savage Defiance is a great second wind, and a completely instant cast at that. I'm going with Improved Defiance and Improved Frenzy, myself. But you should really just do what you want and have fun! 

[voad1]

I will... try too

 

 

I love this game to pieces but it's almost an OCD thing were I must re roll. Thank you for all the information Matt.

[Kaylon]

For people interested in ranged weapons - the Speed bonus (x1.2 attack speed) you can find on some of them does absolutely nothing (tested with Hold-Wall and The Rain of Godagh Field).

[Mysh]

 

Has someone checked using fraps if the modifiers stack additively for the reloading part of ranged weapons? I'm asking because when you look into Cheatengine they are multiplied (1.2x1.5x1.2 for a ranger with gunner, swift aim and chanter buff). 

OK, I tested myself using fraps/virtualdub how the reload time was calculated and everything is multiplicative like I suspected.

Here are the numbers for the arquebus (used the reload time without the delay you have between actions - which is about 5frames):

- 10dex, no other buff - 188frames (base)

- 10dex, swift aim (1.5 x reload time) - 126frames (188/1.5=125.3)

- 10dex, swift aim (1.5 x reload time), gunner (1.2 x reload time) - 105frames (188/1.5/1.2=104.4)

- 20dex, swift aim (1.5 x reload time), gunner (1.2 x reload time) - 81frames (188/1.5/1.2/1.3=80.3)

- 20dex, swift aim (1.5 x reload time), gunner (1.2 x reload time), chant (1.2 x reload time) - 66frames (188/1.5/1.2/1.3/1.2=66.95)

The results are pretty spot on, with the differences within 1frame.

 

 

 

I may have time to work on this again tomorrow. Could you please give me the full formula including the base value, so I can update my calculations?

[Mysh]

 

Things are not simple, for sure. I continually get different peaks for different conditions. I am currently confident though that dexterity builds get the edge on late game, while might builds yield higher DPS early. Let's keep some doubts though, till we verify the formulas.

 

Personally, I've got no doubts - damage modifiers can only continue to go up as the series progresses, which only helps Dex over Might. It's a certainty to me that at some point - whether it be in the PoE 1 expansion(s) or in PoE 2 - Might will be completely and totally overshadowed by Dex for all auto-attacker builds. Casters obviously still benefit from Might to get more out of each spell.

 

Looks like you've done some good work though - though I haven't checked your formulas over in detail. Looking forward to seeing what you can come up with. If you can produce some good graphs for communicating how the current system makes Might less and less relevant as the levels get higher, maybe we can convince Obsidian to change it - hopefully by just making Might multiplicative (the most obvious solution) or making weapon enchants actually change the base damage (the least intrusive and most intuitive solution).

 

 

This is exactly what happens. Yes, I have graphs and I can show you if you like. High might is good early on and then it decays and dexterity easily takes over if you stack savage attack, vulnerable attack and a superb weapon mod.

 

What people also tend to miss is that in cases if high DR, it is also better to have a fast hitting rate, because you still get 20% of damage going through anyway. So the high might may be advantageous in a very narrow window of relative high DR, but not too high. If you chech the foe DR distributions I plotted, you will notice there is no foe at all in this narrow window.

 

So overall, stack savage attack and vulnerable attack all the time. Savage attack is a no brainer for all the time, vulnerable attack needs to be switched off if enemies have very low DR.

 

So, I am waiting feedback on:

1. updating the attack speed formulas

2. on having a volunteer to go through my codes